The tumor stroma is an integral part of the tumor that becomes reprogrammed by unknown mechanisms to co-evolve with epithelial tumor cells and provide an environment conducive for tumor initiation and progression. However, whether alterations in the stroma contribute to the diversity, heterogeneity and plasticity of breast cancer phenotypes, and to the therapeutic responses and clinical outcome in patients is unknown. Our combined preliminary data using mouse models suggests that the Pten and p53 pathways play key and distinct tumor suppressor roles in stromal fibroblasts of the mammary gland. In parallel, gene expression profiling of human breast tumor stroma has identified distinct subclasses of stroma that can predict patient outcomes independent of established clinical variables and the established molecular tumor subtypes. Comparison of gene signatures from Pten-null and p53-null mouse fibroblasts to human tumor stroma signatures show that the mouse signatures are present in human tumor stroma, can help stratify stromal subclasses associated with predominantly non-luminal breast cancers, and can predict poor patient outcome. The overall hypothesis for this proposal is that defining the pathways that regulate tumor-stroma crosstalk will lead to better stratification of breast cancer patients, better informed application of current therapies to patients that are most likely to benefit from them, and to the development of new therapies that target the tumor-stromal interactions uncovered by our studies. Our combined mouse-human approach will utilize novel genetic, genomic and proteomic technologies developed by our groups to 1) expose how stromal cells use integrated transcriptional programs to communicate between themselves and with tumor cells during the initial stages of malignancy, and 2) to identify stromal expression profiles that can better stratify breast cancer patients, predict clinical outcome and be used for serum-based diagnostics. The synergistic efforts between the three projects, utilizing a combined mouse-human approach, is only effective because of the unique expertise and technologies provided by each group with a focus on the single task of understanding the tumor-stroma dialogue in human breast cancer, and in translating such basic information to stratify stroma subclasses, predict clinical outcome and develop specific treatments for individual patients.